The design is a comprehensive mesh network of wireless indoor sensors of various types designed to sense room occupancy and current activities. Based on the gathered information, smart lights would be controlled appropriately. The system would use machine learning to correlate activities and occupancy to lighting scenes.
Many sensor and lighting devices exist on the market that can be the building blocks of the system. Existing solutions require the users to write rules for triggering actions and scenes, and there is no known system that aggregates many various types of sensors into controlling the lighting. The types of sensors that would benefit the system would be door open, motion, IR occupancy, brightness, outlet power usage, weight, passive wifi/bluetooth radio, and "trip wire" sensors.
By combining all the sensor values, it would be easy for machine learning to correlate an activity, such as watching a movie at night to a dim light scene, or reading a book requiring in the recliner to require adequate lighting in that location. The lighting could follow people through the house without manual control. Existing solutions typically rely on one sensor, such as a simple on/off motion sensor, which can result in poor prediction of actions. The proposed system could easily detect when a user is moving in the direction of a light that must be turned on by using input from many different sensors.
Many of these IoT sensors exist on the market with open protocols, such as Zigbee and Z-Wave. Furthermore, many smarthomes already contain at least one sensor, but it's not utilized effectively because it's on a rule based system that leads to a poor user experience. By using machine learning, the proposed system is scalable to the users' budgets, meaning many people can benefit from the system. Automated lighting is the next step in home convenience.
The hardware requirements of the system are fairly low. It needs enough CPU power to process the data in realtime, which should be satisfied by an affordable ARM CPU (such as a Raspberry PI) for medium smarthomes. The device needs to be able to receive the appropriate wireless communications, which would require a Zigbee or Z-Wave radio. It would also need to be connected to the home network, which could be via wifi or Ethernet. For low volume production, the device could be a Raspberry PI with a couple modules in a plastic case. For high volume production, a single-board computer with the necessary radios would be quite feasible.